Gaussian Process Regression for Empirical Radar Cross Section Modeling Based on the SCATR ISAR Dataset

Empirical modeling of Radar Cross Section (RCS) is crucial in defence applications, particularly for guiding the design of specialized Ship Detection Modes (SDMs) tailored for small vessels. This paper presents a novel Gaussian Process Regression (GPR) model for RCS prediction of ships in Inverse Synthetic Aperture Radar (ISAR) imagery. Leveraging 586 Motion Compensated (MO-COMP) ISAR frames representing five ship classes from the real-world Small Craft Automatic Target Recognition (SCATR) dataset, three variants of (i) baseline Multiple Linear Regression (MLR) model, and (ii) GPR model, are constructed. The demonstrated efficacy of the GPR-based approach surpasses that of the baseline MLR model, emphasizing its significant improvement in RCS modeling.